Agentive Modals∗

Matthew Mandelkern, Ginger Schultheis and David Boylan

December 16, 2016

Abstract

We propose a new theory of agentive modals: ability modals and their duals, compulsion modals.After criticizing existing approaches − the existential quantificational analysis, the universal quantifica-tional analysis, and the conditional analysis−we present a new account that builds on both the existentialand conditional analyses. On our account, the act conditional analysis, a sentence like ‘John can swimacross the river’ says that there is some practically available action (in a sense we make precise) whichis such that, if John tries to do it, he swims across the river. We argue that the act conditional analysisavoids the problems faced by existing accounts of agentive modality, and show how it can be extended toan account of generic agentive modal claims. The upshot is a new vantage point on the role of agentivemodal ascriptions in practical discourse: ability ascriptions serve as a kind of hypothetical guarantee, andcompulsion ascriptions as a kind of non-hypothetical guarantee.

Keywords: ability; compulsion; conditional analysis of ability; natural language modality; semantics; prag-matics

1 Introduction

Ability modals are modals like those used in (1), (2), and (3): modals that can be paraphrased with the

dedicated ability constructions ‘able’ or ‘has the ability/capacity’.

(1) John can go swimming this evening.

(2) Mary can touch her nose with her tongue.

(3) Louise is able to pick Roger up from work today.

∗We thank audiences at MIT, at the Semantics and Philosophy in Europe Eighth Colloquium at the University of Cambridge,an anonymous reviewer, and editors for this journal. This paper grew out of earlier work (Mandelkern et al., 2015) presented at the20th Amsterdam Colloquium, and we thank audiences at that presentation as well as three anonymous reviewers for the AmsterdamColloquium. We thank, finally, Alex Byrne, Fabrizio Cariani, Kai von Fintel, Martin Hackl, Samia Hesni, Sabine Iatridou, AbbyEverett Jaques, Matthias Jenny, Justin Khoo, Joshua Knobe, John Maier, Eliot Michaelson, Milo Phillips-Brown, Paolo Santorio,Kieran Setiya, Bradford Skow, Robert Stalnaker, Zoltan Szabo, Timothy Williamson, and Stephen Yablo for valuable discussionand feedback.

1

The modal constructions in (4), (5), and (6) are examples of the duals of ability modals, which we call

compulsion modals:

(4) Lara cannot but eat another cookie right now.

(5) I have to sneeze right now.

(6) I cannot not climb mountains.

We call the class of ability and compulsion modals, taken together, agentive modals. In this paper, we give

a new theory of agentive modals.

We begin by discussing three existing accounts of agentive modals: the orthodox existential account,

according to which a sentence like ‘John can go swimming this evening’ means that John goes swimming

this evening is compatible with a contextually determined set of worlds; the universal account, which holds

that ‘John can go swimming this evening’ means that the proposition John goes swimming this evening is

entailed by a contextually determined set of worlds; and the conditional account, which holds that ‘John can

go swimming this evening’ means that if John tried to go swimming this evening, he would succeed.

We argue that none of these accounts is satisfactory. Along the way, we enumerate two desiderata for

a satisfactory account of agentive modality. First, it must make the right predictions about the neglected

duals of ability modals, compulsion modals (the dual of an operator O is a term which has the meaning of

p¬O¬q). This follows from the broader methodological point that a semantic theory for a certain expression

must make plausible predictions not only about that expression’s meaning when it occurs unembedded, but

also about its meaning when it is embedded under operators like negation. Second, any satisfactory account

of agentive modals must capture the hypothetical nature of ability ascriptions − ability ascriptions tell us

what an agent would do under various alternative circumstances. But it must do so while avoiding a class of

potent counterexamples to the conditional analysis.

In the course of our discussion, we identify two new classes of modals: the duals of ability modals

(compulsion modals); and the class which contains both ability and compulsion modals (agentive modals).

We suggest that these classes play key roles in a broad range of philosophical debates.

Our account, the act conditional analysis, builds on the central insights of the orthodox analysis and the

conditional analysis. We treat ability ascriptions as a kind of hypothetical guarantee. When someone says

‘John can go swimming this evening’, she is informing her interlocutors that going swimming this evening

2

is, in a certain sense, within John’s control. And we treat compulsion modals as a kind of non-hypothetical

guarantee. When someone says ‘I cannot but eat another cookie’, she is saying that refraining from eating

another cookie is not an option for her: she is guaranteed to eat another one.

We elucidate these ideas by arguing that a sentence like ‘John can go swimming this evening’ means

that there is some practically available action (in a sense we make precise) such that if John tried to do

that action, he would go swimming this evening. This account posits a central connection between agentive

modals and two kinds of facts: conditional facts and epistemic facts. Whether an agent is said to be able to

do (or refrain from doing) something depends on what would happen in some (possibly non-actual) scenario

in which she tries to do some relevant action. It also depends on whether we judge that the agent knows how,

in some sense, to perform the action, since we argue that whether an action counts as practically available

− and thus within the domain of quantification for agentive modals − partly depends on whether the agent

knows that it is a way of doing the modal’s prejacent.1 Whether we treat an action as having this status

depends, in part, on whether we have an objective or subjective description of the agent’s practical situation

in mind. We thus distinguish between objective and subjective readings of agentive modals, a distinction

which parallels the distinction between objective and subjective readings of deontic modals.

In the first part of the paper we focus on specific agentive modal ascriptions: agentive modal ascriptions

which have as their prejacent a specific action − an action indexed to a specific time − as in (1), (3), (4),

and (5). We conclude the paper by enriching our account with a generic operator to make sense of generic

agentive modal ascriptions − modal ascriptions like those in (2) and (6), which have as their prejacent a

generic action, one not tied to a specific time. We show that our enriched account makes sense of the fact

that many agentive modal claims say something general about what an agent is equipped to do (or refrain

from doing) in normal circumstances.

2 The Orthodox Account

We start by discussing the orthodox account of natural language modals. This account appears to success-

fully model a variety of “flavors” of modality − deontic, epistemic, circumstantial.2 But we argue that this

account does not make sense of the case we are interested in: agentive modals. In particular, it runs into

1The prejacent of a modal is ambiguously the clause the modal embeds, or the denotation of that clause.2Though a variety of challenges have been leveled at the application of this account to deontic and epistemic modals in recent

years.

3

trouble when it comes to ability modals.

The orthodox account, as applied to agentive modals, traces back to Hilpinen (1969), is taken up in

Lewis (1976), and spelled out in Kratzer (1977, 1981) (who we follow in our presentation). The orthodox

account treats agentive ‘can’ as an existential quantifier whose domain is a set of worlds supplied by context,

in particular by a contextually supplied modal base hc and ordering source gc. Both of these are functions

from worlds to sets of propositions. Given a world w, they together determine a set of “best” worlds relative

to hc and gc, BESThc,gc,w.3 Then:4

(7) Orthodox Account:

JS can ϕKc,w = 1 iff ∃w′ ∈ BESThc,gc,w : JϕKc,w′ = 1.

Informally, pS can ϕq is true just in case JϕKc is true at some “best” world; in other words, just in case JϕKc

is compatible with some contextually salient set of worlds.

On the standard implementation of this account for agentive modals, the modal base is circumstantial,

and the ordering source takes each world to a set of propositions that “holds fixed certain intrinsic features

of the agent in question” at that world (Vetter (2013, 7); see also Portner (2009)). But the resulting truth-

conditions are too weak.5 To see this, suppose Jim and Jo are at a crucial stage in a game of darts and Jo

needs to hit a bullseye to win the game. Jo’s overzealous young child Susie exclaims (8):

(8) Let me take your turn! I can hit the bullseye on this throw.

Susie hardly ever even hits the dartboard when she throws a dart, and she has never gotten a dart to stick.

Intuitively, then, (8) is not true in this scenario.6 We do not say that (8) is clearly false; its negation, ‘Susie

cannot hit the bullseye’, does not strike us as clearly true either (a fact which our theory will try to make

sense of by adverting to indeterminacy). We do say, however, that (8) is not clearly true. But, on the approach

just sketched, (8) is predicted to be clearly, determinately true, since it is certainly compatible with Susie’s3Specifically, let �g,w preorder

⋂h(w) as follows: x �g,w y just in case {ψ ∈ g(w) : y ∈ ψ} ⊆ {ψ ∈ g(w) : x ∈ ψ}. Then

let BESTh,g,w be the set of worlds in⋂h(w) that are not strictly outranked by any other worlds, according to�g,w. For simplicity,

we assume there is such a set (the limit assumption; see Lewis (1981), Stalnaker (1981) for discussion). Relaxing this assumptiondoes not help with the difficulties for the existential analysis; Portner (2009) suggests making ‘can’ a good possibility modal, whichdiverges from the present proposal if we do relax the limit assumption, but this makes the dual too weak: it predicts that pS cannotbut ϕq is true if (say) there is an alternating descending infinite sequence of JϕKc and J¬ϕKc worlds among the best worlds.

4JϕKc,w is ϕ’s extension relative to context c and world w; JϕKc its intension. Unless otherwise noted, we treat ‘can’ and ‘able’interchangeably and intend them to be interpreted agentively throughout.

5We will assume throughout what follows that the modal base is circumstantial, and only consider changes to the orderingsource; changing the modal base won’t help with the problems we raise. Kenny (1976) gives a different, influential critique of theorthodoxy, which we discuss in Section 6.3.

6To fix intuitions, imagine that Susie does throw the dart, and that it falls far short of the dartboard.

4

intrinsic properties, along with local circumstances, that she hit a bullseye.

Note that (8) does have a true reading, which can be paraphrased as (9):

(9) It can happen that I hit the bullseye on this throw.

This is a circumstantial (or “metaphysical”) reading of ‘can’. The present proposal adequately accounts for

this reading, but not for the prominent agentive reading of (8), brought out by (10) rather than (9).

(10) I’m able to hit the bullseye on this throw.7

On the most natural way of spelling it out, then, the orthodox account provides a suitable analysis of

circumstantial modals, but not of ability modals.

A natural first reaction to this case is to limit our domain of quantification to normal worlds, by including

in the value of the ordering source for a world w a set of propositions describing what is normal at w. But

this won’t solve our problem. First, it’s not clear that normality helps even in this case: although hitting the

bullseye is unlikely, it is not obviously abnormal. But suppose we grant that it is abnormal, in a relevant

sense, for Susie to hit the bullseye. But now the proposal overgenerates: agents are often able to do things

which are highly abnormal in just this sense. For example, Susie is a competent speaker of English, and thus

is able to utter the sentence (11):

(11) The world is everything that is the case.

But, being a small child and non-philosopher, circumstances in which she utters this sentence are, intuitively,

at least as abnormal as ones in which she hits the bullseye. The present proposal would thus wrongly predict

that (12) is false:

(12) Susie can now utter, ‘The world is everything that is the case’.

But (12) is clearly, uncontroversially true. Incorporating normality into the ordering source will thus help

only if we can spell out a notion of normality that treats Susie’s hitting a bullseye as abnormal, but Susie’s

uttering (11) as normal. We don’t see a natural way to walk this line.

7A generalization seems to be lurking behind this test: in the semantic/syntactic sense, agentive modals are control modals,whereas circumstantial modals are raising modals. In the discussion of the conditional analysis and our own analysis, we willassume that the prejacent of agentive modals denotes an action, rather than a proposition; this assumption is made for concreteness,and does not have substantial import for either view: we do not mean to take a stand on the underlying question of the semantics ofcontrol (on which see, among others, Chierchia (1989), Brennan (1993), Wurmbrand (2002)).

5

A different approach has abilities themselves, rather than facts about normality, determine what worlds

count as best.8 For instance, we could include JϕKc in the value of the ordering source at w just in case pS

can ϕq is true at 〈c, w〉. But this doesn’t help with our problems: as long as hitting the bullseye is consistent

with Susie’s abilities, we predict that (8) is true. Suppose Susie has k (specific) abilities, of which at most

n can be realized in any given world. Suppose further that for some n of her abilities, realizing all of them,

together with local circumstances, is compatible with her hitting a bullseye (we could easily flesh out our

case so that this holds). In that case, some best world will be one where Susie hits a bullseye, and so we will

predict that (8) is true. Even if we let abilities directly determine the accessibility relation, then, the orthodox

framework still can not make the right predictions.

The orthodox semantics thus does not provide a fruitful framework for analyzing ability ascriptions. We

suggest that this is because it analyzes them in terms of compatibility, an overly weak notion.9

3 The Universal Account and Compulsion Modals

In response to worries like these, some have suggested that agentive ‘can’ has universal, rather than existen-

tial, force.10 We don’t think such an account can work. To see why, we’ll begin with a brief excursus on the

dual of agentive ‘can’.

The dual of ‘can’ has been largely neglected in the literature. Indeed, some have claimed that agentive

‘can’ has no dual.11 A weak interpretation of this claim says that there is no dedicated lexical dual of ‘can’,

a term uniquely interpreted as the dual of agentive ‘can’. This seems true in English − all of the expressions

which express the dual of agentive ‘can’ can also express other flavors of modality − but this fact is not

particularly interesting for our purposes, nor is it unique to agentive ‘can’ (it is also true for other flavors

of modality). A strong interpretation of the claim says that there is no way to express the semantic dual of

agentive ‘can’. This, however, is clearly false. As with any other modal operator, we can form the dual of

‘can’ by putting a negation above and below it; ‘cannot but’ and ‘cannot not’ (italics indicating stress), as in

(13) and (14), realize this semantic pattern, and so are both duals of ‘can’:12

8Thanks to John Maier for suggesting this approach.9This isn’t to say there is no way for the orthodox approach to account for ability modals; see the conclusion.

10Among others, see Kenny (1976, 139), Brown (1988), Giannakidou (2001), Giannakidou and Staraki (2012).11See Hackl (1998). The two versions of this claim have not, as far as we know, been carefully distinguished in the literature,

which may have led in part to the neglect of the dual of ‘can’.12Thanks to Stephen Yablo for pointing out the first of these.

6

(13) Ginger cannot but eat another cookie right now.

(14) Ginger cannot not eat another cookie right now.

‘Must’ and ‘have to’ can also have the meaning of the dual of ‘can’, as in (15) and (16), which seem to be

equivalent to a paraphrase with ‘cannot but’:

(15) I have to sneeze. (Kratzer, 1977)

(16) I must climb K2 this summer.

We believe that data like these have been neglected in part because of an infelicity of nomenclature.

‘Cannot but’ is not an ability modal. (13) does not say that Ginger is able to do something, but that she is

compelled to do something. We thus propose to call modals with the meaning of ‘cannot but’ compulsion

modals. This difference in meaning, however, doesn’t show that ‘cannot but’ is not the dual of ‘can’; rather,

it illustrates the point that duals do not have the same meaning as each other. Compare deontic ‘may’, a

permission modal. Its dual, deontic ‘must’, expresses obligation. ‘May’ and ‘must’ have different meanings,

but the same subject matter: together, they constitute a larger unified class, the class of deontic modals.

Likewise, agentive ‘can’ and its dual have different meanings but the same subject matter: they too belong

to a larger unified class, which we call agentive modals.13 Carefully delimiting this class is important for the

study of natural language and of traditional philosophical problems. For instance, we suspect that the strong

necessity modals that appear in anankastic conditionals are best analyzed as compulsion modals; likewise

for the modals adverted to in philosophical discussions of freedom and practical necessity.14

Once we have the dual of ‘can’ clearly in view, we can quickly dispense with the universal analysis of

‘can’. On that analysis, ‘cannot but’ will have existential force. This can’t be squared with the intuition that

‘cannot but’ expresses compulsion: we do not see how we could make sense of this intuition if ‘cannot but’

had existential force.15 The universal analysis might seem tenable when we only consider intuitions about

unembedded ability ascriptions, but not when we turn our attention to ability ascriptions embedded under

negation.

13See, among others, Belnap (1991), Maier (2015b) for related but distinct usages.14On the former, see, for example, von Fintel and Iatridou (2005); on the latter, see Aristotle (2002), Frankfurt (1969), Williams

(1982), Setiya (2007).15Brown (1988) takes this prediction on board, claiming, implausibly, that the dual of ‘can’ is ‘might’. Brown’s approach is

formally similar to the conditional analysis, except that it does not make Stalnaker’s uniqueness assumption. The present points tellagainst an approach like that, but a more sophisticated version is worth exploring, such as one that adopts a homogeneity constraintlike that of von Fintel (1997).

7

4 The Conditional Analysis

A more promising approach than either the orthodox or universal approach treats pS can ϕq as meaning,

roughly, pS would ϕ if S tried to ϕq.16 We argue that this approach avoids the problems raised for the

orthodox analysis, but faces a variety of problems of its own.

The conditional analysis borrows the selection function from Stalnaker (1968)’s theory of conditionals:

a contextually supplied function fc(ψ,w) from a proposition ψ and world w to the “closest” world where

ψ is true. We’ll use italics to indicate the proposition which fc takes as an argument. Let ϕ range over

strings denoting actions, like ‘going swimming at 2 pm’. We will model actions as properties − functions

from individuals to propositions − and assume they are specific and indexed to a time, rather than generic

actions. Let Jϕ(S)Kc be the proposition that S does JϕKc. Then:

(17) Conditional Analysis:

JS can ϕKc,w = 1 iff Jϕ(S)Kc,fc(S tries to ϕ,w) = 1.

On this approach, pS can ϕq is true just in case S does JϕKc in the closest world where S tries to do JϕKc;

informally, just in case S would do JϕKc if she tried to.17,18

The conditional analysis (henceforth ‘CA’) does better than the orthodox existential account. Recall

Susie and the bullseye. Consider the conditional (18):

(18) If Susie tried to hit the bullseye now, she would.

Intuitively, (18) is not clearly true. That said, it does not strike us as clearly false either. We will follow

Stalnaker (1981) in saying that the actual state of affairs, together with semantic facts, does not suffice to

decide whether a conditional like this is true or false: a conditional like (18) is neither determinately true nor

determinately false, but rather indeterminate.19 Crucially, the CA predicts that (19) shares the truth value of

16This is an old philosophical idea, traceable to Hume (1748), taken up by Moore (1912) among others; for formulations in amodel-theoretic framework, see especially Lehrer (1976), Cross (1986), and Thomason (2005).

17This last paraphrase depends on adopting Stalnaker (1968)’s theory of the conditional. For those who do not adopt it, thisparaphrase (and similar ones below) should be taken with a grain of salt.

18Trying plays a key role in both the conditional analysis and our own analysis. We will not give a semantics for ‘try’ here(on which see, for example, Sharvit (2003) and citations therein). Because both the conditional analysis and our account analyzeability ascriptions in terms of trying and conditional reasoning, they make certain predictions about the relative rate of acquisitionof certain concepts. For instance, these accounts predict that subjects will not be able to reason about abilities until they can reasonabout trying. This hypothesis is, as far as we know, compatible with the present state of empirical research, which suggests thatchildren acquire the concept of trying, or purposive action in general, at a very early age; see, among others, Woodward (1998).Thanks to John Maier and Jonathan Phillips for discussion.

19A different approach, equally suitable for our purposes, treats (18) as probably false (Hawthorne, 2005).

8

(18).

(19) Susie can hit the bullseye now.

If (18) is not determinately true or determinately false, the CA predicts that (19) is not, either. This explains

why we are reluctant to accept (19), as well as its negation in (20):

(20) Susie can’t hit the bullseye now.

The CA also rightly predicts that (12) is clearly true, since (22) is clearly true.

(12) Susie can now utter ‘The world is everything that is the case’.

(21) If Susie now tried to utter ‘The world is everything that is the case’, she would.

Unlike the orthodox account, the CA is thus able to distinguish examples like (19) from examples like (12).

It does this by capturing the fact that ability requires something more than compatibility: an ability is a kind

of hypothetical necessity, a guarantee (or something close to it) that you will get something done if you try.

Despite its promise, however, the CA has a number of serious problems, which we turn to now.

4.1 Compulsion Modals

First, the CA makes the wrong predictions about compulsion modals. On the CA, pS cannot but ϕq will be

predicted to mean pNot (If S tries to not ϕ, she does not ϕ)q. On Stalnaker’s theory of conditionals, this is

equivalent to pIf S tries to not ϕ, she does ϕq. Thus:

(22) Compulsion Modals (Conditional Analysis):

JS cannot but ϕKc,w = 1 iff Jϕ(S)Kc,fc(S tries to not ϕ,w) = 1.

In words: pS cannot but ϕq is true just in case the closest world where S tries to not JϕKc is one where S

does JϕKc; put differently, just in case if S tries not to JϕKc, S does JϕKc.

But this is too weak. pS cannot but ϕq means not only that S does JϕKc if she tries not to, but that she

does JϕKc no matter what she tries to do. If compulsion modals meant only what the CA says they do, then

pS cannot but ϕq and pS cannot but not ϕqwould be consistent. For example, ‘Ginger cannot but eat another

cookie’ and ‘Ginger cannot but not eat another cookie’ would be consistent. This strikes us as false.

9

To bring this out, consider the following case.20 Suppose, unbeknownst to him, the buttons of the elevator

in John’s building have been incorrectly wired: If John presses the button marked ‘1’ it takes him to the

basement, and if he presses the button marked ‘B’ it takes him to the ground floor (suppose these are the

only two buttons). John is in the elevator, and will press one of the two buttons. The CA predicts that (23) is

true:

(23) John cannot but go to the first floor.

This is because, if John tries to not go to the first floor, then he will hit ‘B’, and thus end up on the first floor.

For similar reasons, the CA predicts that (24) is also true:

(24) John cannot but go to the basement.

Both these predictions are wrong. (23) is false: if John presses button ‘1’, he will get to the basement. (24)

is also false, for similar reasons. Most strikingly, (23) and (24) are clearly inconsistent.

4.2 Ability Modals

The second problem for the CA is that, in a broad range of cases, intuitions about ability ascriptions come

apart from intuitions about the corresponding conditional.

Consider first cases in which pS can ϕq is intuitively false, but the corresponding conditional true.

Suppose, for example, that John is planning to go to a movie.21 Ann invites John to dinner, and he replies:

(25) I’m sorry, I’m not able to go; I’m going to a movie.

There is a prominent true reading of (25) − John can’t make it to dinner because he’s going to a movie. But

(26) is clearly true:

(26) If John tried to have dinner with Ann, he would succeed.

The CA says that (25) is false if (26) is true, so it cannot predict the true reading of (25).

Importantly, there are a number of reasons to think that ‘able’ really is agentive in (25), rather than a

deontic or bouletic modal. First, ‘able’ has a default agentive reading: out of the blue, speakers are inclined

to hear ‘able’ as being about the agent’s abilities, rather than about circumstantial, epistemic or deontic

20Thanks to Robert Stalnaker for suggesting a similar case to us.21This case is adapted from Thomason (2005).

10

possibility. Second, we can easily set up the case so that John has no deontic commitment to go to the

movie: simply suppose he is going by himself because he just wants to see the movie. Third, the retraction

data associated with (25) are not what we would expect from a deontic or bouletic modal. When someone

makes a deontic or bouletic claim, rejoining with a claim about ability feels like a non-sequitur. But in

the present case, pointing out that John really does have an ability to go to dinner feels like a natural and

effective response to (25), as in (27):

(27) Of course you’re able to meet me − just skip the movie and come to dinner!

Examples like (25) are widespread; speakers often use pI can’t ϕq or pI’m not able to ϕq to communicate

that they have a commitment which conflicts with doing JϕKc. A semantics for ‘can’ must capture this fact.

The CA does not.

A similar case, where the relevant conditional seems true and the corresponding ability ascription false,

comes from Lehrer (1968).22 Suppose that Larry is offered a bowl of red candy. He has a pathological phobia

of red candy; nothing could induce him to take such a candy. (28) seems false.

(28) Larry is able to take a candy.

But, one might think, if he tried to take the candy, he’d succeed − the closest world where he tries to take

the candy is one where he does not have the phobia in question. If that’s right, the CA wrongly predicts that

(28) is true. (We are not completely convinced that (29) is actually true in this situation.

(29) If Larry tried to take the candy, he would.

After all, Larry has a phobia! See Albritton (1985) for attempts to dismiss cases like this along these lines.

But we are happy to grant the judgment that (29) is true. If this is right, it’s another problem for the CA

which, as we’ll show, our account does not have; if not, examples like (25) already provide decisive reason

to reject the CA.)

The examples we have discussed so far are counterexamples to the CA as a sufficient condition for the

truth of an ability ascription. There are also counterexamples to the CA as a necessary condition for the truth

of ability ascriptions: cases in which pS can ϕq is true but pS would ϕ if she triedq is false. Consider the

following case, adapted from Austin (1961). Jones is a skilled golfer with an easy shot onto the green. Matt

22See also Chisholm (1964); a variant has a comatose agent Len. We want to say that ‘Len can take a walk’ is false; but if Lentried to take a walk, he would be conscious, and so he would in fact take a walk.

11

says:

(30) Jones is able to make this shot right now.

Matt has said something true. Now suppose Jones takes the shot and misses the green. We may still judge

Matt to have said something true; afterwards, we can truly say (31):

(31) Jones (was able to/had the ability to) make that shot at that time.23

Yet given how the selection function is defined, the closest world where Jones tries to make the shot is the

actual world, since Jones in fact tried to make the shot, and any world is more similar to itself than to any

other world. Since Jones actually misses, the CA wrongly predicts that (30) is false. A way to press this

point is to point out that sentences like (32) are often felicitous:

(32) Jones is able to make this shot right now, though if he tries, he of course might miss.

The CA predicts that the two conjuncts in (32) are not jointly assertable, since pIf ϕ, then ψq and pIf ϕ, then

might not ϕq are not jointly assertable on any plausible theory of the conditional.

A final kind of case that poses problems for the CA is one in which an agent can do something, but only

if she does not try to do it, as in (33):24

(33) David can breathe normally for the next five minutes.

(33) is true if David is healthy. But if David tried to breathe normally, he’d concentrate on his breathing and

end up breathing abnormally.

From a technical point of view, all of the cases in this section are easy to respond to: We can simply

choose a selection function for ability ascriptions that selects worlds in a way that matches our intuitions.25

The problem with this response is that it uncouples the CA from the analysis of conditionals, and thus from

our intuitions about conditionals and similarity in general.26 Without an intuitive characterization of the

23There is also a false reading of (31), brought out when ‘was able to’ has perfective aspect (see Bhatt (1999) and others),but all that matters for our purposes is that there is a true reading, which is clearly brought out in languages that mark aspectmorphologically, when ‘was able to’ has imperfective aspect. Thanks to Nilanjan Das and Raphael Turcotte for data in Hindi,Bengali, and French.

24See Vranas (2010) for discussion. Again, not everyone accepts these judgments. We find this case compelling, but our rejectionof the CA does not depend on these judgments.

25Thomason (2005) suggests a response along these lines.26For instance, to make sense of cases like (30), we would have to say that the actual world is not the most similar world to

itself (even if we moved to a Lewisian analysis with multiple most similar worlds, we would have to say that the actual world is notamong the most similar worlds to itself).

12

altered selection function, the resulting theory is neither predictive nor explanatory.

5 The Act Conditional Analysis

Many have thought cases like those discussed in the last subsection refute the CA.27 We agree. But we think

that the CA is on the right track. It rightly captures the hypothetical nature of abilities: whether you are able

to perform a particular action depends in some way on what happens under relevant alternate circumstances.

We give an account of agentive modals that preserves this insight, yet avoids the CA’s problems.

Our account incorporates into the meaning of ‘can’ a layer of quantification over a contextually supplied

set of actions. Let AS,c,w be a set of actions which are − in a sense to be precisified − practically available

to an agent S in a context c and world w. Where, again, fc is Stalnaker’s selection function and ϕ denotes a

specific action:

(34) Act Conditional Analysis:

JS can ϕKc,w = 1 iff ∃A ∈ AS,c,w : Jϕ(S)Kc,fc(S tries to A,w) = 1.

According to the act conditional analysis (henceforth ‘ACA’), pS can ϕq is true just in case there is some

practically available actionA such that the closest world where S tries to doA is a world where S does JϕKc;

in other words, just in case there is some practically available action such that if S tries to do it, she does

JϕKc. Note that that action does not need to be JϕKc: this is the key difference between our account and the

CA.28,29

At a first pass, we may assume that in many cases, in an ability ascription with the form pS can ϕq,

then JϕKc ∈ AS,c,w. In those cases the predictions of the ACA come very close to those of the CA. For

example, if we make this assumption in evaluating (35), then the action going for a swim in the pool tonight

is included in ALouis,c,w.

(35) Louis can go for a swim in the pool tonight.

27See, among others, Austin (1961), Lehrer (1968), van Inwagen (1983).28Chisholm (1964) makes a similar suggestion. As far as we know his suggestion hasn’t been taken up in the subsequent

literature, perhaps because he himself sketches a fairly serious objection to it; the ACA, however, avoids that objection by restrictingthe set of actions quantified over. See also Albritton (1985), Setiya (2007), and Maier (2015b) for discussions with some connectionto the present proposal.

29We leave open whether we need to encode a non-accidental connection between S and JϕKc in order for pS can ϕq to comeout true. We are inclined to think that the connection is implicated rather than encoded.

13

In this case, (35) is true if Louis will go swimming tonight should he try to do so.

The similarity of the ACA to the CA allows it to inherit the main virtue of the CA sketched above: it is

able to distinguish between cases like (19) and (12), repeated here:

(19) Susie can hit the bullseye now.

(12) Susie can now utter ‘The world is everything that is the case’.

Assuming for simplicity thatASusie,c,w includes just the actions throw a dart towards the bullseye and utter

‘The world is everything that is the case’, and their complements, we predict that (12) is clearly true, since

if Susie tries to utter ‘The world is everything that is the case’, she succeeds. On the other hand, we predict

that (19) is not clearly, determinately true, since none of these actions is determinately such that if Susie

tries to do it, she succeeds.

As we will now show, in addition to capturing these attractive predictions of the CA (and thus avoiding

our objections to the orthodox account), the ACA avoids the problems for the CA raised above.

5.1 Compulsion Modals

First, our approach makes plausible predictions about the meaning of compulsion modals:

(36) Compulsion Modals (Act Conditional Analysis):

JS cannot but ϕKc,w=1 iff ∀A ∈ AS,c,w :Jϕ(S)Kc,fc(S tries to A,w)=1.

Informally: For every action A practically available to S in c, S does JϕKc in the closest world in which S

tries to do A. In other words, no matter what S tries to do (among the actions we are treating as practically

available in c), S ends up doing JϕKc. This looks like a plausible prediction − much more plausible than the

CA’s prediction that pS cannot but ϕq is true just if S does JϕKc provided S tries not to. Consider (13) again:

(13) Ginger cannot but eat another cookie right now.

Intuitively, (13) says that Ginger is compelled to eat another cookie: no matter what she tries to do, she’ll

eat another one. This is precisely what we predict. Moreover, the ACA rightly predicts that pS cannot but

ϕq and pS cannot but not ϕq are inconsistent:30 if S does JϕKc no matter what she tries, then it’s not the

30Provided that AS,c,w is non-empty, a condition we assume is met in most cases.

14

case that S does J¬ϕKc no matter what she tries. Thus, for instance, (23) and (24), repeated here, will be

inconsistent:

(23) John cannot but go to the first floor.

(24) John cannot but go to the basement.

Note that the plausibility of our predictions here stems in part from our choice to use the selection

function from Stalnaker (1968), which selects a single world; had we followed Lewis (1973) in adopting a

selection function which selects a set of worlds instead of a single world, we would make the implausibly

weak prediction that pS cannot but ϕq is true just in case, for every practically available action, if S tries to

do that action, then she might do JϕKc.

5.2 Ability Modals

In addition to making plausible predictions about compulsion modals, the ACA also makes plausible pre-

dictions about ability modals, avoiding the counterexamples to the CA discussed in Section 4.2. Consider

first cases in which pS can ϕq is intuitively false, even though it is true that S would JϕKc if S tried. Recall

John, who says (25):

(25) I’m sorry, I’m not able to go [to dinner]; I’m going to a movie.

Unlike the CA, the ACA can predict that (25) has a prominent true reading. To be sure, if John tried to go

to dinner, he’d succeed. But, on our proposal, this does not guarantee that ‘John can go to dinner’ is true:

the action meeting Ann for dinner (or something much like it) must also be treated as practically available

to John. If this condition is not met, then there is no action in AJohn,c,w such that trying to do it guarantees

that John meets Ann for dinner. In that case, we predict that ‘John can go to dinner’ is false, and that (25) is

true.

We conjecture that meeting Ann for dinner is not treated as practically available in this context because

John has decided against this action. One way to test the plausibility of this hypothesis is to see whether

insisting on the availability of the action meeting Ann for dinner can modulate intuitions, since in general,

speakers tend to defer to an insistence on a larger domain of quantification. Suppose Ann replies:

(27) Of course you’re able to meet me − just skip the movie and come to dinner!

15

It seems that Ann has said something true; after hearing (27), we are inclined to judge that it is true that John

can meet Ann for dinner. By uttering (27), Ann ensures that AJohn,c,w include meeting Ann for dinner, and

so (27) comes out true.

In a moment we will discuss, in more disciplined and general terms, what practical availability amounts

to. For the present, though, we will show how flexibility in how this parameter is determined lets us respond

in a similar way to the other cases discussed in Section 4.2.

Recall the case in which Larry is offered a bowl of candy, but his phobia prevents him from taking one.

If Larry is so phobic that we cannot even entertain the possibility of his trying to take the candy, then we

may well not treat taking the candy as practically available for Larry. In that case, we will predict that ‘Larry

can take the candy’ is false, even if ‘If Larry tried to take the candy, he would’ is true.

We can make similar moves in response to cases where pS can ϕq is true, even though it is false that S

would ϕ if S tried. Recall the golf case. We said that (30) has a true reading.

(30) Jones is able make this shot right now.

Now suppose Jones aimed to the left of a certain tree; because of a freak gust of wind, the ball was blown off

course. Had he tried to aim to the right, he would have made the shot. Let aiming to the right be inAJones,c,w.

Then we predict that (36) is true even though Jones actually misses. This nicely captures intuitions: (37) is

a natural way to describe the case.

(37) Well, he was able to make the shot; all he had to do was aim to the right.

We often ascribe abilities to agents even when they are not certain to succeed at a given action should they

try, and even in cases where they fail when they in fact try. The ACA nicely accounts for these uncertain

abilities; we discuss cases like this further in Section 6.2.

Finally, appeal to AS,c,w lets the ACA explain why (33) and its ilk strike us as true.

(33) David can breathe normally for the next five minutes.

There is something relevant (we may suppose) such that if David tries to do that, he breathes normally (say,

working on a paper), and thus the ACA predicts that (37) is true.

16

5.3 Practical Availability

We have shown that, by incorporating quantification over a set of actions, the ACA can avoid our counterex-

amples to the CA, without uncoupling the analysis of agentive modals from intuitions about conditional

facts. But unless we say more about how this set of actions is determined − that is, about what practical

availability amounts to − we face a charge of ad hoc maneuvering. In this section we give a general charac-

terization of practical availability that regiments the intuitions elicited in the last section. We do not aim to

provide a universal characterization of practical availability; we believe it is a genuinely context-dependent

notion, and as such may be determined in a variety of different ways in different contexts. But we hope to

say enough to give a sense of speakers’ default way of thinking about practical availability, and thus to make

the predictions of our account more concrete.

Consider the following case, suggested by Timothy Williamson, which will help bring out the constraints

we need to place on practical availability. Imagine a grid with one hundred buttons, labelled 1-100. For

exactly one of these buttons, if Lizzie pushes it, she wins a prize, but she doesn’t know which one it is. She

gets one shot at winning the prize. It seems clear that (38) has a prominent agentive reading on which it’s

not (determinately) true; after all, Lizzie doesn’t know which button to push in order to win the prize.

(38) Lizzie is able to win the prize.

This judgment is important to get clearly in mind; one way to bring it out is to imagine Lizzie in a high-stakes

situation, say in a rapidly descending plane. One of one hundred buttons on the console turns on autopilot;

the others ignite the fuel. The panicked crew-members ask:

(39) Can anyone here turn on autopilot?

Suppose Lizzie replied with (40):

(40) I can!

Then the crew-members would rightly hold her to account if they learned that she doesn’t know which

button enables autopilot.

On their most prominent readings, then, neither (40) nor (38) is determinately true. (We do not hold that

they are determinately false either; they may be indeterminate. Further, as we discuss in a moment, there are

also true readings of these which can be brought out in certain ways.)

17

Which actions should count as practically available in order for us to predict these judgments? Consider

again (38). We want to say that ALizzie,c,w contains the actions winning the prize and not winning the prize.

If just those actions are practically available, we predict that (38) is not determinately true, since there will

be no practically available action which is determinately such that if Lizzie tried to do it, she would win the

prize. But ALizzie,c,w should not include the more specific actions pushing button one, pushing button two,

and so on. IfALizzie,c,w did contain these more specific actions, then we’d predict that (38) is true. Suppose

the seventh button is the winning one; it’s clearly true of Lizzie that if she tried to push the seventh button,

she would succeed, and so there would be a practically available action such that if she tries to do it, she

wins.

But how do we predict that this is howALizzie,c,w is set? This question looks particularly pressing when

we compare our judgments about (38) to our judgments about (41).

(41) Lizzie can press the kth button.

It seems clear that every instance of (41) is true, for 1 ≤ k ≤ 100. In order for these to be true, however,

something like pressing the kth button must be treated as practically available, for every k. But then we will

predict that (38) is true, since there will be some practically available action such that if Lizzie tries to do

that, she wins.

We can make the right predictions about both (38) and (41) by treatingAS,c,w as sensitive to the prejacent

of the modal (which is, after all, part of the contextually available information). In particular, we suggest that

an action ψ will typically count as practically available for S in a context c just in case S could reasonably

conclude in favor of doing ψ with respect to the goal of doing the prejacent or its complement.31 This is

not meant as a strict rule; actions might count as practically available even when they are not reasonable in

this sense (for instance, after the fact, the action that an agent actually does will almost invariably count as

practically available, whether or not it was reasonable). But we believe reasonableness provides a helpful

general heuristic for fleshing out the set of practically available actions.

Reasonableness will involve many things. Among them will be a certain kind of epistemic standing.

We propose that an action A is reasonable in the relevant sense only if the agent knows that A is a way to

bring about the prejacent − or knows that it is a way to bring about its complement − relative to a certain

31We include the complement as a goal in order to make correct predictions about negated ability modals.

18

description of her practical situation (more on this below).32 In other words, the agent must know that, given

that description, if she does A, then she does the prejacent; or that if she does A, she does the prejacent’s

complement.

Applying this idea to our case: Relative to the goals of winning the prize, and not winning the prize −

the prejacent of (38) and its complement − and assuming that we take Lizzie’s knowledge of her practical

situation to provide the relevant description of her practical situation, the set of practically available actions

will just be winning the prize, plus any actions that Lizzie knows are ways of not winning the prize. Impor-

tantly, an action like pressing the seventh button will not be treated as practically available, since Lizzie does

not know that this is a way to win the prize (or to not win the prize). By contrast, she of course does know

that winning the prize is a way of winning the prize, so, assuming that this action is otherwise a reasonable

thing for Lizzie to conclude in favor of, it will be treated as practically available. It follows that (38) will

not be determinately true, since no practically available action is determinately such that if Lizzie tries to

do it, she wins the prize. By contrast, relative to the action pushing the kth button (the goal made salient by

the prejacent of (41)), pushing the kth button will itself generally be treated as practically available, since

Lizzie will generally know this is a way of pushing the kth button. So we predict that (41) will be true in its

context.

Our account thus lets us predict a true reading of (41) and an indeterminate reading of (38). So far so

good. But in addition to an indeterminate reading, (38) also seems to have a true reading. This is easiest to

bring out if we consider a bystander commenting on the situation; the bystander could say (42):

(42) Lizzie can win the prize; she just has to push the seventh button.

How can we predict this? We have said that an action counts as practically available only if the agent knows

that it is a way of bringing about the prejacent relative to a given description of her practical situation.

There are different ways of supplying the relevant description. On the subjective way of thinking about it

we have assumed just now, the description of Lizzie’s practical situation is limited to what she knows about

her actual situation. We think this reading is the default reading; it can be brought out by focusing on what

makes sense given an agent’s limited information, as in (43):

(43) Lizzie can’t put the plane on autopilot − she has no idea which button to press!

32Perhaps (true) belief would suffice; we do not have strong committments on this point.

19

But on a more objective way of thinking about it, the description of Lizzie’s practical situation is just a

complete description of all the facts relevant to her actual situation. An objective reading is elicited by

making salient the actual facts at hand, as in (42).

The epistemic constraint on practical availability can thus be read either objectively or subjectively.

Somewhat surprisingly, then, as for deontic modals, we can distinguish objective from subjective readings

for agentive modals.33

The present characterization of practical availability helps answer a possible worry about our view.

Suppose again that the seventh button is the winning button. On our view, ‘can’ is upward monotone: If ϕ

entails ψ, then pS can ϕq entails pS can ψq.34 Since pushing the seventh button entails winning the prize in

this context, we predict that (44) entails (38) (repeated here). Thus the falsity of (38) should entail that (44)

is false, too.

(44) Lizzie can push the seventh button on the grid.

(38) Lizzie is able to win the prize.

But (44) sounds clearly true in nearly any context. This is a prima facie puzzle for any upward monotone

account, like ours, and indeed, as an editor for this journal points out, for any account which validates the

substitution of equivalent prejacents under agentive modals − since in this case, pushing the seventh button

is equivalent to winning the prize.35

Our characterization of practical availability can resolve this puzzle, however. Though we do predict that

(44) is sometimes false (whenever (38) is), it does not follow that we can ever hear a false reading of (44).

On our account, false instances of (44) are elusive in the sense of Lewis (1996): Any context in which (38)

is not true is one in which (44) is not true, but the moment we entertain (44), the context changes in a way

that makes (44) come out true, since by changing the prejacent, we change the set of practically available

actions.

33Our structural approach to the distinction parallels Cariani et al. (2013)’s approach to deontic modals. We assume this distinc-tion admits of gradations tracking different ways of specifying a practical situation.

34This generally matches intuitions, modulo “free choice effects” (Kamp, 1973) which must be accounted for; more thoroughdiscussion of monotonicity must await future work.

35To be precise, what we need to get the problem going is intersubstitutability of contextual equivalents, or upward contextualentailment. Both differ from upward logical entailment, which is all that really follows from our view. But, under the standardassumption that the relevant selection function always selects a world in the context set when there is one which verifies its propo-sitional argument (Stalnaker, 1968), then both those principles will indeed follow from our view whenever, for every practicallyavailable action, it is compatible with the context that the agent in question tries to do that action − a condition which seems to bemet in the present case, and indeed in many cases.

20

A point in favor of this approach is that it looks like we need upward monotonicity − or at least the

substitutability of contextual equivalents − to explain certain reasoning patterns. Suppose Matt says that

Lizzie is unable to win the prize. Out of the blue, we’re inclined to agree with him. But now suppose that

Sally responds as follows:

(45) Lizzie can push the first button, the second, and so on. But one of these is the winning button. So

she can push the winning button, and thus she’s able to win!

It would be reasonable for Matt to retract his claim after hearing Sally’s case. Our account captures this fact,

since we predict that Sally’s inference in (45) is valid, while also making sense of the difference in intuitions

about the truth of (38) versus (44).

We conclude this section by showing that our characterization of practical availability regiments and

explains the assumptions we made in showing how the ACA avoids the counterexamples to the CA. Recall

John, who has decided to go to a movie. Is there an action he could reasonably conclude in favor of doing,

relative to the pair of goals going to dinner with Ann, not going to dinner with Ann, such that if he tries to do

it, he goes to dinner with Ann? It is easy to get the intuition that there is not: since he has decided to go to

a movie, it would not be rational for him to deliberatively conclude in favor of something inconsistent with

that. People should stick with their decisions. On the other hand, one can also get the intuition (with slightly

more difficulty) that there is a reasonable way for him to go to dinner with Ann. After all, John could simply

cancel his movie plans; people should be willing to revisit decisions. Depending how we come down on

this, we will treat going to dinner with Ann as practically available (in the latter case) or not (in the former),

and thus predict both true and false readings of (46).

(46) John can go to dinner with Ann.

We can say similar things about phobic Larry. Taking candy would not be a reasonable course of action

for phobic Larry to deliberatively conclude in favor of, for obvious reasons, and thus will not be treated as

practically available.

One might worry about circularity in this account of practical availability. Among other things, facts

about S’s abilities might influence what counts as a reasonable course of deliberation; indeed, this seems

to the case here. But, like most projects of semantic analysis, our aim is elucidation, not reductive analysis.

This discussion also brings out an important fact about our characterization of practical availability, worth

21

emphasizing here: The epistemic constraint we discussed above is only meant to be a necessary condition on

practical availability, not a sufficient one. In addition to the epistemic constraint, more general considerations

about reasonableness will also influence whether an action is treated as practically available.

Let’s now turn to Jones the golfer. Suppose that Jones did not know that aiming just to the right of the

tree would result in a successful shot, because he did not know that a gust of wind would arise at just the

moment he made the swing. Jones did know, however, that, given such a gust of wind, aiming just to the

right of the tree would result in a successful shot. So, relative to an objective description of his practical

situation, aiming just to the right of the tree is practically available. Assuming that he would have made the

shot had he tried to aim just to the right of the tree, we thus predict a true, objective reading of (30).

(30) Jones is able to make this shot right now.

This objective reading can, again, be brought out by emphasizing the facts of Jones’s situation − facts he

might not be aware of − as in (37), repeated here:

(37) Well, he was able to make the shot; all he had to do was aim to the right.

By contrast, for a less skilled golfer, Milo, the analogue of (30) will not be true, even on an objective reading.

For even if Milo knew all of the relevant facts about his practical situation, it is likely that, if Milo tried to

aim to the right, he would still not make the shot. Thus the truth of (30) depends in an important way on the

fact that Jones is a skilled golfer.

One might wonder whether the distinction between a subjective and objective reading of agentive modals

which we have spelled out in this section, and which we are relying on in our analyses of these cases, is really

necessary. An alternate hypothesis is that what we have been calling an objective reading of an agentive

modal is really just a circumstantial modal.36 We think, however, that there is an important distinction to

be made here: while the objective agentive reading may well entail the circumstantial reading in most cases

(since it makes salient a world where the prejacent happens), the converse is not true.

To see this point, take a variant on the Susie case. Susie is now playing against Smith, an expert dart

player who is extremely reliable in making bullseyes at this distance. What they don’t know is that they are

playing with darts that are slightly heavier than usual; to make a bullseye, they would have to compensate

for this.37 Jo, who is watching the game, says:36Thanks to an anonymous reviewer for this journal for pushing us on this question.37Smith’s predicament is similar to that of Jones the golfer.

22

(47) Smith is able to hit a bullseye now − he just has to compensate for the weight.

This sounds perfectly true here. This is what we have been calling an objective reading of ‘able’; the subjec-

tive reading is false, since Smith doesn’t know that he needs to compensate for the weight of the bullseye.

By contrast, (48) does not sound (determinately) true; the situation here is identical to the original case

involving Susie.

(48) Susie is able to get a bullseye now − she just has to compensate for the weight.

But paraphrases eliciting the circumstantial reading pattern differently:

(49) It can happen that Susie hits the bullseye.

(50) It can happen that Smith hits the bullseye.

Both (49) and (50) sound true. Thus it is circumstantially possible both for Smith and for Susie to hit a

bullseye. But if the prominent true reading of (47) really is a circumstantial reading, as the present objection

goes, then we should expect a prominent true reading for (48). This is not borne out. Nor can a subjective

agentive reading predict this contrast between (47) and (48), since, again, (47) is false on a subjective agen-

tive reading. By contrast, the objective reading of agentive modals which we have posited predicts precisely

the contrast we observe here. Since Smith is highly skilled, there is something which (determinately) would

have resulted in him making the shot, if he had tried it − he simply does not in this case know what action

that is. Not so for Susie: since she is a small child, no action is (determinately) such that, had she tried

it, it would have resulted in her hitting the bullseye. Our account is thus right to distinguish circumstantial

readings from objective agentive readings.

5.4 Interim Summary

This completes our exposition of the ACA. The ACA builds on the orthodox existential account (insofar as

it incorporates existential quantification) and on the CA (insofar as it incorporates a conditional element),

but avoids the problems we raised for both accounts. Putting the pieces together: our account says that pS

can ϕq is true just in case there is some practically available action A such that if S tries to do A, S does

JϕKc. Practical availability is, in turn, a context-dependent notion. We have proposed the following rough

heuristic for thinking about it: an action A is practically available to an agent S with respect to a salient goal

23

ψ just in case S could reasonably conclude in favor of doing A in order to do ψ. When it comes to ability

ascriptions we assume that the salient goals are typically the prejacent of the ascription and its complement,

and we assume, inter alia, that S can reasonably conclude in favor of doing A in order to do ψ only if she

knows that − given a certain description of the world − if she does A, she does ψ. Our characterization of

practical availability is somewhat vague.38 But so are ability ascriptions, and the vagueness of the two seem

to go hand in hand.

Our account answers to an important intuition about what ability ascriptions are used to do. When

someone says pI can ϕq, she is assuring her interlocutors that JϕKc is within her control in a certain way.

This is why we do not judge that Susie can hit the bullseye, or that Lizzie can turn on autopilot. This is

precisely what the ACA predicts: pS can ϕq says that there is some relevant action − some action that S

could reasonably deliberate about and conclude in favor of doing − such that if S tries to do it, S succeeds

in doing JϕKc.

Ability ascriptions, on our account, serve as a kind of hypothetical guarantee. This is a strong prediction.

Two things are important to keep in mind here, however. First, we distinguish between an objective and

subjective reading of agentive modals. On an objective reading, if pS can ϕq is true, then for some practically

available action A, if S tries to do A, she does JϕKc; but it may nonetheless be the case that she does not

know which action A has this status, and so the ability ascription will not have the practical effect of any

kind of guarantee. Second, as we discuss in a moment, generic ability ascriptions do not guarantee that JϕKc

is within S’s control, but only something weaker: that JϕKc is in her control if conditions are normal. With

these caveats in mind, we believe that the characterization of an ability ascription as a kind of hypothetical

guarantee does indeed match intuitions, and casts a new light on the role of ability ascriptions in practical

dialogue.

Our account also answers to an important intuition about what compulsion ascriptions are used to do:

When someone says pI cannot but ϕq, she is saying that she will do JϕKc no matter what she tries to do.

In other words, compulsion ascriptions serve as a kind of non-hypothetical guarantee. We believe that this

matches intuitions about the meaning of claims like this, and their role in practical dialogue.

One upshot of our account is that we predict that ability ascriptions of the form pS can try to ϕq are

quite weak. An ability ascription of this form simply says that there is something practically available such

38This characterization− including its inherent vagueness−may help explain the impurity of ability ascriptions, in the sense ofKnobe and Szabo (2013); considerations about deontic matters, for instance, influence ability ascriptions insofar as they influencewhat we count as reasonable and thus practically available. See Mandelkern and Phillips (2016) for further discussion of this idea.

24

that if S tries to do it, she tries to do JϕKc. Thus, for instance, as long as JϕKc is practically available, pS can

try to ϕq will be predicted to be true. This prediction seems right to us.

Does our account make trying too easy? It predicts that pS can try to ϕq is consistent with pS cannot

ϕq, since the latter says that for any practically available action, if S tries it, she does not do JϕKc; and the

former says that for some practically available action, if S tries to do it, she tries to do JϕKc.39 This may look

wrong; note that (51) is marked:

(51) I can’t fix the fridge, but I can try to.

The infelicity of (51) is not immediately predicted on our account. But we think our prediction is correct:

being unable to do something is consistent with being able to try to do it. To see this, note that the third-

person variant on (51) in (52) is perfectly coherent:

(52) John can’t fix the fridge; he can only try to.

(53) provides a similar case, using ‘able’ to select for the agentive reading of the modals, and providing

set-up which may make this kind of conjunction more natural:

(53) Anyone’s able to try to quit smoking. But few are actually able to do it, and among those, even

fewer will. [Milo Phillips-Brown (p.c.)]

The coherence of (52) and (53) suggests that pS cannot ϕq is indeed consistent with pS can try to ϕq. We

suspect, then, that the explanation of the infelicity of (51) is pragmatic. The first conjunct of (51) licenses

the inference that the speaker believes that, for any practically available action A, if he tries to do A, then he

will fail to fix the fridge. There is something in bad faith about trying to do ψ if you believe that you can’t do

ψ. Why would you be offering to try, if you knew you would fail?40 (All this is compatible, of course, with

S trying to do ψ in good faith even though S cannot do ψ, as long as S does not believe this about herself.)

And indeed, note that (51) is improved when this pragmatic constraint is waived − say, if the speaker is

being pressured to try to fix the fridge. Since no pragmatic story of this kind will extend to third-personal

variants of (51), the explanation we’ve offered makes sense of the contrast between (51) and (52) within our

framework: both are semantically consistent, as our account predicts, but (51) is (usually) infelicitous for

39We are grateful to an anonymous reviewer for this journal for raising this issue, and for suggesting an example like (51), aswell as the possibility of a pragmatic response.

40There is thus an analogy between sentences like (51) and Moore sentences.

25

pragmatic reasons.

We note, finally, that tying practical availability to an agent’s epistemic state commits us to a certain

position on the relation between know-how and ability. We will not offer an analysis of know-how here,

but presumably knowing how to do A is generally closely connected to knowing of some B that doing B

guarantees doing A.41 On our account, knowing of some B that it has this status does not guarantee that

one is able to do A; one must also be such that, for some such B, if one tries to do B, one does A. But

being able to do A does guarantee, inter alia, that there is some action B that one knows has this status

(relative to some description of one’s practical situation) − and, moreover, that if one tried to do B, one

would do A. This strikes us as a plausible hypothesis about the relationship between ability ascriptions and

know-how. Know-how does not guarantee ability, since ability is not a purely cognitive matter. But it is a

partly cognitive matter: ability does entail know-how.

6 Genericity

The ACA is an account of specific agentive modal ascriptions. In this section we discuss how to generalize

the ACA to generic agentive modal ascriptions like (2): agentive modal ascriptions whose prejacent is an

action not indexed to a specific time (either explicitly or implicitly).

(2) Mary can touch her nose with her tongue.

We then discuss a variety of related issues involving genericity and agentive modals.

6.1 Generic Agentive Modal Ascriptions

We propose that generic agentive modal ascriptions are simply specific agentive modal ascriptions under-

neath a generic operator.42 A specific agentive modal ascription has as its prejacent a specific, time-indexed

41Not everyone agrees this is always true; see, among others, Brownstein and Michaelson (2016). See Stanley (2011) andcitations therein for more on this topic.

42See Bhatt (1999), Hacquard (2010), Maier (2015a) for similar ideas, implemented in importantly different ways. For instance,Hacquard gives agentive modals a standard existential quantificational analysis, and then proposes that the fact that their meaningis felt to be stronger than predicted on that analysis stems from an implicit generic operator licensed by imperfective aspect. Thisapproach faces a worry from the fact that the ability modals seem to have a stronger meaning than the existential account wouldpredict even when they are in contexts which seem not to have a generic operator. For instance, ‘Yesterday at 3 p.m., Susie wasable to hit a bullseye’ (or its corollary in languages with overt aspect, with ‘was able to’ in the imperfective) seems to us not tobe clearly true, as spoken about Susie the child dart-player, and thus to have a meaning stronger than that predicted by a bareexistential analysis. But this sentence does not seem to have an implicit generic operator: generic operators do not happily coexistwith temporal phrases like ‘yesterday at 3 p.m.’ (to see this, note that adding ‘generally’, which is usually licensed when a covertgeneric operator is in play, renders this sentence infelicitous; note also that if we replace ‘Susie’ with an existentially quantified

26

action, like hitting a bullseye in Somerville at 5 p.m. this Friday, and says of an agent that there is some

practically available action such that if she tries to do it, then she does the prejacent. A generic agentive

modal ascription has as its prejacent a general action − one which is not indexed to a particular time − like

hitting the bullseye, and says of an agent that, in a suitable proportion of normal circumstances, there is an

action practically available in those circumstances such that, if she tries to do it, then she does the prejacent.

In the rest of this section we spell out this idea in more detail, though without going too far into the

vexed question of exactly what a generic operator is;43 we adopt the following implementation primarily for

concreteness. We assume that genericity is due to an operator GEN, usually tacit, with roughly the meaning

of ‘generally’. GEN binds a time variable t in a tensed sentence ψt to yield another sentence which says,

essentially, that ψ is true in enough normal situations, relative to the standards of normalcy determined by

the context and world.44 More formally, with t a free time variable in ψt, ts the time of a situation s, and

‘GEN’ subscripted with the variable it binds:

(54) JGENt(ψt)Kc,w=1 iff in a suitablec proportion of normalc,w situations s, JψtsKc,s = 1.

Roughly, GEN takes an untensed sentence, and returns a sentence that says that the original sentence is true

in enough normal cases, according to the standards of normalcy established by the world of evaluation and

context.

Two things need to be made explicit in our account of agentive modals before we can show how it

interacts with GEN. First, we have left time indices out of our presentation so far for simplicity; we now

need to be explicit about them. Ability ascriptions can contain two distinct time indices, one attached to

the modal (which says when the agent has the ability in question) and one attached to the action (which

says when the action takes place). We will ignore the first of these, assuming for simplicity that it does not

interact with the GEN operator, and denote the second with a subscript t, so that an ability ascription now

has the form of (55), where ϕt denotes a specific action that takes place at t (we change the subject variable

to J to avoid confusion with situation variables).

phrase, like ‘a child’, no generic reading of the existential is available). To handle cases like this, it seems to us that a view alongthese lines would have to advert to some other, further mechanism of strengthening besides an implicit generic operator. Thanks toValentine Hacquard for valuable discussion.

43See Krifka et al. (1995, Chapter 1) for an overview, especially §1.2.6 on the approach we follow. See Sterken (2015), Nickel(2016) and citations therein for more recent discussion.

44It doesn’t matter for our purposes what a situation is; we could think of it as a world, or part of a world, centered on at leasta time, but we assume that sentences can be evaluated for truth not only at a world, but also at situations; and likewise that theconditional selection function fc, and the function AS,c to practically available actions, are defined on situations as well as onworlds.

27

(55) J can ϕt.

Second, we must now represent AS,c as part of the logical form of an ability ascription − that is, as part of

the structure that underlies semantic composition − so that GEN can bind into it, as follows:45

(56) JJ can (AJ,c)(ϕ)Kc,w = 1 iff ∃A ∈ AJ,c,w : Jϕ(S)Kc,fc(J tries to A,w) = 1.

GEN binds the time index on the prejacent. Thus on our account generic ability ascriptions have the form

and truth conditions in (57):

(57) JGENt(J can (AJ,c)(ϕt))Kc,w = 1 iff at a suitablec proportion of normalc,w situations s: ∃A ∈

AJ,c,s : Jϕts(J)Kc,fc(J tries to A,s) = 1.

Informally, a generic ability ascription with the surface form pJ can ϕq says of J that, in enough normal

situations s, there is a practically available action at the time of s which is such that if J tries to do that

action, J brings it about that JϕKc holds of her at that time. In short: In normal situations, J can do JϕKc.

To make these truth-conditions plausible, we must assume that what counts as a contextually normal

situation depends on the prejacent. For example, consider (58):

(58) George can make great ratatouille.

When evaluating (58), the set of normal situations will include only those where George has all the in-

gredients, tools, etc. needed to make ratatouille. This kind of sensitivity is needed for generics in general,

however; for instance, we must say something similar in evaluating (59):

(59) George makes great ratatouille.

In addition to giving intuitive truth conditions for generic ability ascriptions, our account also makes an

interesting further prediction: the possibility of scopal interactions between GEN and other constituents of

ability ascriptions. We believe this prediction is borne out, and explains a well-known ambiguity in ability

ascriptions.46 Consider the following two possible logical forms for a generic ability ascription:

(60) GENt(J can (AJ,c)(ϕt))

45In a fully explicit semantics, we would treat AS,c as a variable at logical form which gets saturated by a contextual variableassignment.

46We are indebted to Justin Khoo for first suggesting this possibility to us.

28

(61) J can (AJ,c)(GENt(ϕt))

We’ve assumed that generic ability ascriptions have the form of (60), and mean that, in enough normal

situations, there is something available for J to do such that if she tries to do it, she does JϕKc. But if GEN

can scope under AJ,c, as in (61), we will find generic ability ascriptions which mean that there is something

(now) available for J to do such that if she tries to do it, then in enough normal circumstances, she will do

JϕKc.47 Both of these two scopal possibilities seem to be manifested. Consider (62):

(62) John can speak Greek.

As Lewis (1976) observes (broadly following Aristotle (1968, 417a)), a sentence like (62) can mean two

things. First, it can mean that John is able, in general, to speak Greek − in other words, that he is a fluent

speaker of Greek. Second, it can mean that John has the physiological and cognitive apparatus to come to

be a fluent speaker of Greek, by contrast, say, with an ape; in other words, that John could learn Greek if

he wanted to. These two glosses correspond precisely to the two logical forms (60) and (61), respectively.

The first parse says that, in most normal situations, there is something practically available which John can

try to do in order to speak Greek. The second parse says there is something practically available such that if

John tries to do it, then, in normal circumstances, he speaks Greek. In short, the first says there is normally

something John can do to speak Greek; the second says there is something John can do to become a normal

Greek-speaker. The possibility of scopal interactions predicted by the present account thus explains this

ambiguity in ability ascriptions.48

6.2 Uncertain Abilities

As we noted in Section 5.2, we often ascribe abilities to agents even when they are not certain to succeed at

a given action should they try. The present discussion of genericity, together with our earlier discussion of

objective versus subjective readings of agentive modals, helps us make sense of this use of ability ascriptions.

We touched on this topic in discussing the golfer Jones, where we said that (30) can have a true reading

even if Jones actually misses, provided there was some knowledge that Jones lacked such that, if he had had

it, he would have made the shot.

47Assuming that free time variables are set by default to the time of utterance.48This explanation of the ambiguity presupposes that it arises only for generic, not specific, ability ascriptions. This, however,

seems to match intuitions.

29

(30) Jones is able to make this shot right now.

But not all uncertain ability ascriptions can be handled by an appeal to lack of knowledge along these lines.

In this section, we show that by carefully distinguishing generic from specific agentive modal claims −

as well as objective from subjective ones − the ACA gives a satisfying account of these uncertain ability

ascriptions.

Suppose Stephen Curry has a difficult three-point shot to make. Consider (63):

(63) Curry can make this shot.

Intuitively, given Curry’s skill, (63) is true in many contexts, even though in many cases − perhaps the

majority − Curry will miss when he tries to make a shot like this. How do we predict this?

In light of the present discussion, we distinguish two readings of (63): a generic reading and a specific

reading. On the generic reading, (63) means something like (64):

(64) He can make this kind of shot.

Even if Curry was not able to make the very shot he tried to make, we can predict that sentences like (64)

are true, if Curry is able to make this kind of shot in a suitable proportion of normal cases. What counts

as a suitable proportion of normal cases is, again, highly context-dependent. If the baseline of successful

shots is quite low, what counts as a suitable proportion may likewise be quite low. Compare a generic like

‘Rats carry bubonic plague’. This may be true even though only a small proportion of rats in fact carry the

plague.49 Thus in a situation that involves making a highly risky shot, it may suffice for the truth of a generic

like (64) that Curry succeeds, say, half the time. A generic interpretation thus may account for many true

readings of uncertain ability ascriptions like (63).

What about specific readings of (63), forced by adding an overt time index, as in (65)?

(65) Curry can make this shot right now.

Our predictions about this case will depend on how we flesh it out. We distinguish cases in which Curry

is ignorant in a relevant way about his practical situation (ignorance cases) from those in which he has no

relevant ignorance (no-ignorance cases).

Begin with ignorance cases. We can treat these cases just as we treated Jones’s golf shot above: In these

49See again Krifka et al. (1995) and citations therein.

30

cases, (65) will have a true objective reading (but not a true subjective reading). Imagine, for example, that

Curry did not know that his teammate was going to set a screen for him. Had he known about the screen, he

would have used it, and made the shot; in fact, he didn’t know about it, and so he went for a different shot

and missed. Then we predict a true objective reading of (65), which might be elicited by saying something

like (66):50

(66) I was able to make that shot; the only reason I didn’t is that I didn’t know about the screen.

Consider now no-ignorance cases − cases in which Curry is not ignorant about any relevant features of

his practical situation. These subdivide into two classes. One kind of case involves a failure of execution.

Imagine, for example, that Curry tried to do a specific action which he knew was a way to make the shot,

but his body failed him − he was tired and so was not quick enough to escape his defender. In a case like

this, our intuition is that Curry was simply not able to make that very shot − he was too tired to execute the

necessary moves. Assuming there is no other action such that, had he tried to do it, he would have made the

shot, then we rightly predict that (65) has no true reading in this scenario.

The second kind of no-ignorance case is one in which Curry’s body does not fail him, but for some

reason he nonetheless fails to execute the specific game-winning action. We can imagine this happening in

cases of inattention or lack of focus. In this case, we are inclined to think there is a true reading of (65):

Curry was able to make the shot; he just had to pay closer attention. Suppose that Curry knew that A was

a way to make the shot, given what he knew about his practical situation. And suppose that Curry tried to

make the shot, but failed. We can still predict a true reading of (65), if we maintain that the action Curry

tried was not A. A may be a highly specific and taxing action to attempt (aiming with this arc and spin

from this position); Curry may have simply failed to try to do that action (instead simply trying something

non-specific like making a shot). A lack of focus, in other words, may involve a failure to try to do the

requisite very specific action. (A common complaint from coaches: ‘You’re not really trying!’.) If that’s the

right diagnosis, we predict a true reading of (65): A is a practically available action for Curry, and if Curry

had tried to do A, he would have made the shot. He just didn’t try to do A.

Uncertain ability ascriptions like those discussed here bring up a variety of complex issues; our distinc-

tions between objective versus subjective − and generic versus specific − agentive modals help make sense

of the subtle intuitions in these cases.

50(65) will also have a false, subjective reading; after all, Curry didn’t know about the screen.

31

6.3 Kenny’s Objection

Our account of generic agentive modals helps answer a classic objection to modal analyses of ability, due

to Kenny (1976). Kenny points out that on a modal analysis like the orthodox one, pS can (ϕ or ψ)q is

predicted to entail p(S can ϕ) or (S can ψ)q. This entailment also follows from the ACA, provided ϕ and ψ

denote specific actions. For suppose that (67) is true:

(67) S can (ϕ or ψ).

Then there is a practically available action A such that the closest world where S tries to do A is one where

she does Jϕ∨ψKc. That world is either one where she does JϕKc or one where she does JψKc. It follows from

our semantics that, if we hold the context fixed, (68) is true.

(68) (S can ϕ) or (S can ψ).

But this entailment is not obviously in line with intuition. (69) does not obviously entail (70):

(69) Leo can, right now, draw a red card or draw a black.

(70) Leo can, right now, draw a red card or Leo can, right now, draw a black card.

We have two things to say in response. Our first response parallels what some defenders of conditional

excluded middle say about purported counterexamples. On this response, it is not determinately true that

Leo can right now draw a red card; nor is it determinately true that Leo can right now draw a black card.

Nonetheless, the disjunction (70) is determinately true, since no matter which way we fill things out, one of

these disjuncts is true− even though we don’t know, and there may be no fact of the matter, which. We think

this response makes good sense of our intuitions about the relation between (69) and (70). When we reflect,

we are inclined to think that (70) follows from (69), even though it may be that neither ‘Leo can draw a red

card’ nor ‘Leo can draw a black card’ is determinately true.

The second response to Kenny’s objection appeals to genericity. We predict that the entailment from

(67) to (68) will be invalid when the ability ascriptions in question are read generically, since (71) obviously

does not entail (72):

(71) GENt(J can (AJ,c)(ϕt ∨ ψt))

(72) GENt(J can (AJ,c)(ϕt))∨GENt(J can (AJ,c)(ψt))

32

It might be the case that, in a suitable proportion of normal situations, J can do Jϕ ∨ ψKc, without it being

the case that, in a suitable proportion of normal situations, J can do JϕKc, or that, in a suitable proportion of

normal situations, J can do JψKc. We thus straightforwardly predict that (73) fails to entail (74):

(73) Leo can (generally) draw a red card, or draw a black card.

(74) Leo can (generally) draw a red card, or Leo can (generally) draw a black card.

6.4 Masks and Finks

Our distinction between specific and generic abilities also helps us respond to another classic objection to

analyzing abilities in terms of conditionals, concerning cases involving so-called masks and finks (see, for

example, Martin (1994), Fara (2008)).

Suppose I am able to break the vase on my desk if I drop it. But suppose the vase is wrapped in bubble

wrap, which I can’t presently remove. The bubble wrap masks my ability to break the vase. So I am able to

break the vase, even though there is no practically available action such that if I try to do it, I break the vase.

This case constitutes a prima facie objection to both the CA and the ACA.

What should we say about this case? We maintain that the specific ability ascription (75) is, in fact, false.

(75) I can break this vase right now by dropping it.

I can’t break the vase right now; the bubble wrap would thwart my attempt to do so. But the mask leaves

intact my generic ability to break the vase:

(76) I can (generally) break the vase.

(76) is true in the case described − under normal circumstances the vase would break if I dropped it, since

it would not be wrapped in bubble wrap. I have a generic ability to break this vase but do not have a specific

ability to break the vase right now, since it is wrapped in bubble wrap. And if I do not have a specific ability

to break the vase, this case poses no counterexample to our account.

We can respond similarly to concerns about finks, which are like masks except that they remove abilities

rather than masking them. Once we clearly distinguish specific from generic ability ascriptions, we see that

finks and masks do not threaten our account.

33

6.5 Trying

Not everything that can be the grammatical subject of an agentive modal ascription can try. This is a prima

facie problem for our account. We close this section by showing how genericity can help with this problem.

On the one hand, we think the notion of trying extends further than one might first imagine; consider for

instance (77):

(77) The flower can follow the sun. [Kieran Setiya (p.c.)]

We think it plausible that we are conceiving of the flower as trying to follow the sun; after all, we might say

that the flower is trying to get as much light as possible. But in other cases it is not plausible to think of the

subject of an ability ascription as trying, as in (78):

(78) This elevator is able to carry three thousand pounds. [Irene Heim (p.c.), attributed to Maria

Bittner]

It doesn’t make sense to say that the elevator tries to carry three thousand pounds. We can dramatize this

point with examples like (79):

(79) This rock can tip the balance. [Martin Hackl (p.c.)]

Rocks do not try to do things. Furthermore, these modals look like ability modals, not circumstantial modals,

since they can be paraphrased as ‘has the ability to’.

How can we accommodate these examples? Here’s a sketch of a response. (78) can be glossed as saying,

roughly, that there is some available action (in this case, loading the elevator with three thousand pounds of

freight and pressing ‘up’) such that in the closest world where the operator of the elevator tries to take that

action, the elevator carries three thousand pounds. Likewise, (79) can be glossed as saying, roughly, that

there is some available action (in this case, putting the rock on the balance) such that in the closest world

where some relevant agent tries to take that action, the rock tips the balance.

More formally, we say that GEN can bind into the subject index on the set of actions and into the first

argument of the selection function, in addition to the time index. Ignoring tense, a sentence like (78) would

thus have the form and meaning:

(80) JGENR(J can (AR,c)(ϕ))Kc,w = 1 iff

34

for a suitablec proportion of normalc,w agents R: ∃A ∈ AR,c,w : Jϕ(J)Kc,fc(R tries to A,w) = 1

Informally: Enough normal agents are such that, in normal situations, there is some practically available

action A such that if they try to do A, they bring it about that J does ϕ. So, for example, (78) will mean

that enough normal agents can make it the case that the elevator carries three thousand pounds right now;

(79) will mean that enough normal agents can make it the case that the rock tips the balance. The ACA

can thus accommodate agentive modals whose grammatical subject is not an agent in any intuitive sense by

appealing to genericity.51

Note that a similar move may be possible in order to account for some ‘-ble’ and ‘-ile’ constructions,

like ‘breakable’.52 Building constructions like these out of ability ascriptions may look implausible; after

all, (81) clearly does not say anything about whether the vase would break if it tried something.

(81) This vase is breakable.

But we may posit an implicit generic agent who does the relevant trying, so that (81) means that, in normal

circumstances, a normal agent can break the vase. -ble and -ile constructions are complex and varied, but

this may provide a basis for an analysis of the subset of these constructions which are, in some broad sense,

about ability.53

7 Conclusion

We have advocated a new semantics for agentive modals, the act conditional analysis. Our semantics builds

on an old idea about ability: that what one can do depends on what would happen if one tried. But our

account avoids the problems that led us, and most other researchers, to reject the conditional analysis. We

achieve this by incorporating quantification over a contextually supplied set of actions into the meaning of

‘can’. This approach makes plausible predictions about agentive modals, avoiding the array of objections

we enumerated for the conditional analysis, as well as the orthodox and universal approaches.

We have highlighted the relation between ability modals and compulsion modals, arguing that making

the right predictions about the meaning of compulsion modals provides a crucial criterion of adequacy for

51A different approach to cases like this would be to make use of a slightly modified notion of “trying” in our semantics, onewhich is like our ordinary notion of trying but can be extended to non-agents. This alternative seems well worth exploring.

52Though certainly not all; for instance, it is hard to see how, for instance, ‘edible’ would be put into this mold.53See Maier (2015c) and Vetter (2014) for recent discussions related to the present proposal.

35

accounts of ability modals. We have further suggested that distinguishing the class of compulsion modals −

and the broader class of agentive modals, which includes both ability and compulsion modals − is critical

for the theory of natural language modality, as well as for an array of broader philosophical issues.

We conclude by mentioning some of those issues, as well as some questions that we have left open.

First, we think that compulsion modals may play a role in making sense of intuitions about freedom and

compulsion in the philosophical debate on free will. Appeal to the notion of practical availability may allow

us to make sense of the malleable and contradictory intuitions about freedom and related notions, as well as

to make precise the traditional idea that freedom, on the one hand, and ability and compulsion, on the other,

are closely connected.

Another important question which we leave open is the cross-modal picture which our account suggests.

On our account, agentive ‘can’ differs in important ways from circumstantial, deontic, and epistemic ‘can’,

at least as those are treated in the orthodox framework. Like the orthodoxy, our semantics treats agentive

‘can’ as an existential quantifier. But in contrast to the orthodoxy, in our account, ‘can’ quantifies over a set

of actions (properties), not a set of worlds; and a conditional selection function plays a crucial role in its

semantics. If we leave unchanged the orthodox account of circumstantial, deontic, and epistemic ‘can’, we

must thus maintain that ‘can’ is lexically polysemous: It can be interpreted either as an ordinary existential

quantifier, to yield a circumstantial, deontic, or epistemic reading; or it can be interpreted as on the ACA, to

yield an agentive reading.

Is this proposed polysemy a substantial drawback? The answer to this will depend partly on the cross-

linguistic picture − how uniformly languages use the same word for an agentive modal which they use for

other kinds of modals− as well as how successful the orthodoxy ends up being for other flavors of modality.

We should note, though, that we can in theory preserve the predictions of the ACA, while preserving the

uniformity of the orthodox framework, by shoehorning the ACA into the orthodox framework. For any

agent S, just let the modal base take any world w to ∅ and let the ordering source take any world w to

{{w′} : ∃A ∈ AS,c,w(w′ = fc(S tries to A,w)}. This account makes the same predictions as ours. But

− assuming that the elements of the semantic formalism are meant to map onto a metasemantic story in

some straightforward way − it involves awkward metasemantic commitments. A metasemantics for both

the ACA and this orthodox recasting of the ACA must account for how AS,c,w and fc are determined, since

both views refer to these parameters. But this recasting of our view in the orthodox framework also makes

reference to a highly gerrymandered ordering source. It is hard to see how this ordering source could play

36

any role in an explanatory metasemantics. So we think there is reason simply to dispense with it, and to

adopt the ACA in the form in which we have spelled it out.

Unification, however, might also be possible in a different direction: not by shoehorning the ACA into the

orthodox framework, but rather by replacing the orthodox framework with a unified framework structured

more like the ACA. Indeed, a number of recent proposals have departed from orthodoxy in a manner similar

to the ACA, incorporating quantification over sets of properties or propositions into the meaning of modals.

This raises the possibility that our proposal could be part of a new, unified approach to modality.54 We leave

further exploration of this question for future work; the answer will turn on further work on the semantics,

metasemantics, and crosslinguistic typology of modals.

A final important question which we leave open is how ability modals interact with tense and aspect to

yield ‘actuality implications’ from pS was able to ϕq to pS did ϕq (see Bhatt (1999) and Hacquard (2006)).

There is more to say about agentive modals, then. However these details are worked out, we hope to

have laid the framework for an account which answers to an important intuition about what agentive modals

are used to do. When someone says pI can ϕq, she is giving a kind of hypothetical guarantee about her

relationship to JϕKc. She is assuring her interlocutors that JϕKc is within her control in a certain way − that

is, that there is some action that she could reasonably conclude in favor of doing, such that if she tries to

do it, she succeeds in doing JϕKc. And when someone says pI cannot but ϕq, she is giving a kind of non-

hypothetical guarantee about her relationship to JϕKc. She is assuring her interlocutors that refraining from

JϕKc is not an option for her; that is, that she will JϕKc no matter what.

ReferencesAlbritton, R. (1985). Freedom of Will and Freedom of Action. In Proceedings and Addresses of the American Philosophical

Association, volume 59, pages 239–251. American Philosophical Association.Aristotle (1968). De Anima (On the Soul). Clarendon Press.Aristotle (2002). Nicomachean Ethics. Oxford University Press, Oxford.Austin, J. L. (1961). Ifs and Cans. In Philosophical Papers, pages 151–180. Oxford University Press, London.Belnap, N. (1991). Backwards and Forwards in the Modal Logic of Agency. Philosophy and Phenomenological Research,

51(4):777–807.Bhatt, R. (1999). Ability Modals and Their Actuality Entailments. In Shahin, K., Blake, S., and Kim, E.-S., editors, WCCFL 17

Proceedings, volume 17, pages 74–87.Brennan, V. M. (1993). Root and Epistemic Modal Auxiliary Verbs. PhD thesis, University of Massachusetts Amherst.Brown, M. A. (1988). On the Logic of Ability. Journal of Philosophical Logic, 17:1–26.Brownstein, M. and Michaelson, E. (2016). Doing without Believing: Intellectualism, Knowledge-how, and Belief-Attribution.

Synthese, 193(2815):1–22.Cariani, F. (2013). ‘Ought’ and Resolution Semantics. Nous, 47(3):534–558.

54For recent work along structurally similar lines, see, among others, Cariani et al. (2013) and Cariani (2013) for related ap-proaches to deontic modals; Moss (2015) on epistemic modals; Yalcin (2012) and Willer (2013) on attitude verbs; and Villalta(2008) on verbs of desire.

37

Cariani, F., Kaufmann, M., and Kaufmann, S. (2013). Deliberative Modality under Epistemic Uncertainty. Linguistics and Philos-ophy, 36:225–259.

Chierchia, G. (1989). Anaphora and Attitudes De Se. In Bartsch, R., van Benthem, J., and van Emde Boas, P., editors, Semanticsand Contextual Expression, pages 1–31. Dordrecht Reidel.

Chisholm, R. M. (1964). J. L. Austin’s Philosophical Papers. Mind, 73(289):1–26.Cross, C. B. (1986). ‘Can’ and the Logic of Ability. Philosophical Studies, 50(1):53–64.Fara, M. (2008). Masked Abilities and Compatibilism. Mind, 117(468):843–865.von Fintel, K. (1997). Bare Plurals, Bare Conditionals, and Only. Journal of Semantics, 14:1–56.von Fintel, K. and Iatridou, S. (2005). What to Do if You Want to Go to Harlem: Anankastic Conditionals and Related Matters.

MS, MIT.Frankfurt, H. G. (1969). Alternate Possibilities and Moral Responsibility. The Journal of Philosophy, 66(23):829–839.Giannakidou, A. (2001). The Meaning of Free Choice. Linguistics and Philosophy, 24:659–735.Giannakidou, A. and Staraki, E. (2012). Ability, Action, and Causation: From Pure Ability to Force. In Mari, A., Beyssade, C., and

Prete, F. D., editors, Genericity. Oxford University Press.Hackl, M. (1998). On the Semantics of “Ability Attributions”. Master’s thesis, MIT.Hacquard, V. (2006). Aspects of Modality. PhD thesis, MIT.Hacquard, V. (2010). Ability Modals: A ♦ in the Generic Rough. Slides from the Dispositions, Abilities and States Workshop.Hawthorne, J. (2005). Chance and Counterfactuals. Philosophy and Phenomenological Research, 70(2):396–405.Hilpinen, R. (1969). An Analysis of Relativised Modalities. In Davis, J., Hockney, D., and Wilson, W., editors, Philosophical

Logic, volume 20. Springer.Hume, D. (1748). An Enquiry Concerning Human Understanding. Oxford University Press, Oxford, beauchamp edition.Kamp, H. (1973). Free Choice Permission. In Proceedings of the Aristotelian Society, volume 74, pages 57–74.Kenny, A. (1976). Human Abilities and Dynamic Modalities. In Manninen, J. and Tuomela, R., editors, Essays on Explanation and

Understanding, pages 209–232. D. Reidel.Knobe, J. and Szabo, Z. G. (2013). Modals with a Taste of the Deontic. Semantics and Pragmatics, 6(1):1–42.Kratzer, A. (1977). What ‘Must’ and ‘Can’ Must and Can Mean. Linguistics and Philosophy, 1(337-355).Kratzer, A. (1981). The Notional Category of Modality. In Eikmeyer, H. and Rieser, H., editors, Words, Worlds, and Contexts: New

Approaches in Word Semantics. de Gruyter.Krifka, M., Pelletier, F. J., Carlson, G. N., ter Meulen, A., Link, G., and Chierchia, G. (1995). Genericity: An Introduction. In

Carlson, G. N. and Pelletier, F. J., editors, The Generics Book, pages 1–124. University of Chicago Press, Chicago.Lehrer, K. (1968). Cans without Ifs. Analysis, 29(1):29–32.Lehrer, K. (1976). ‘Can’ in Theory and Practice: A Possible Worlds Analysis. In Brand, M. and Walton, D., editors, Action Theory,

pages 241–270. D. Reidel.Lewis, D. (1973). Counterfactuals. Harvard University Press, Cambridge, MA.Lewis, D. (1976). The Paradoxes of Time Travel. American Philosophical Quarterly, (145-152).Lewis, D. (1981). Ordering Semantics and Premise Semantics for Counterfactuals. Journal of Philosophical Logic, 10(2):217–234.Lewis, D. (1996). Elusive Knowledge. Australasian Journal of Philosophy, 74(4):549–567.Maier, J. (2015a). Ability, Modality, and Genericity. Manuscript.Maier, J. (2015b). The Agentive Modalities. Philosophy and Phenomenological Research, 40(1):113–134.Maier, J. (2015c). Dispositions and Ergativity. The Philosophical Quarterly, 65(260):381–395.Mandelkern, M. and Phillips, J. (2016). ‘Force’: Topicalization, Context-Sensitivity, and Morality. MS.Mandelkern, M., Schultheis, G., and Boylan, D. (2015). I believe I Can ϕ. In Brochhagen, T., Roelofsen, F., and Theiler, N., editors,

20th Amsterdam Colloquium, pages 256–265.Martin, C. (1994). Dispositions and Conditionals. The Philosophical Quarterly, 44:1–8.Moore, G. (1912). Ethics. Williams and Norgate, London.Moss, S. (2015). On the Semantics and Pragmatics of Epistemic Vocabulary. Semantics and Pragmatics, 8(5):1–81.Nickel, B. (2016). Between Logic and the World: An Integrated Theory of Generics. Oxford University Press.Portner, P. (2009). Modality. Oxford University Press, New York.Setiya, K. (2007). Cognitivism about Instrumental Reason. Ethics, 117:649–673.Sharvit, Y. (2003). Trying to be Progressive: The Extensionality of Try. Journal of Semantics, 20(4):403–445.Stalnaker, R. (1968). A Theory of Conditionals. In Rescher, N., editor, Studies in Logical Theory, pages 98–112. Oxford: Blackwell.Stalnaker, R. (1981). A Defense of Conditional Excluded Middle. In Harper, W. L., Stalnaker, R., and Pearce, G., editors, Ifs:

Conditionals, Beliefs, Decision, Chance, and Time, pages 87–105. D. Reidel.Stanley, J. (2011). Know-How. Oxford University Press.Sterken, R. K. (2015). Generics in Context. Philosophers’ Imprint, 15(21):1–30.Thomason, R. H. (2005). Ability, Action, and Context. MS.van Inwagen, P. (1983). An Essay on Free Will. Oxford University Press, New York.Vetter, B. (2013). ‘Can’ without Possible Worlds: Semantics for Anti-Humeans. Philosophers’ Imprint, 13(16):1–27.Vetter, B. (2014). Dispositions without Conditionals. Mind, 123(489):129–156.

38

Villalta, E. (2008). Mood and Gradability: An Investigation of the Subjunctive Mood in Spanish. Linguistics and Philosophy,31:467–522.

Vranas, P. (2010). What Time Travelers May Be Able to Do. Philosophical Studies, 150(1):115–121.Willer, M. (2013). Dynamics of Epistemic Modality. Philosophical Review, 122(1):45–92.Williams, B. (1982). Practical Necessity. In Moral Luck, pages 124–131. Cambridge University Press.Woodward, A. L. (1998). Infants Selectively Encode the Goal Object of an Actor’s Reach. Cognition, 69(1):1–34.Wurmbrand, S. (2002). Syntactic vs. Semantic Control. In Zwart, J.-W. and Abraham, W., editors, Studies in Comparative Germanic

Syntax, pages 95–129.Yalcin, S. (2012). Context Probabilism. In Aloni, M., Kimmelman, V., Roelofsen, F., Sassoon, G. W., Schulz, K., and Westera, M.,

editors, Proceedings of the 18th Amsterdam Colloquium, pages 12–21.

39